胜利的大会：本地 AI 会议同事，录音→深度业务纪要→知识库，macOS 首发

by Mycelium Protocol

---

GitHub：zhangchunquan298-anhui/shengli-dahui
产品名：胜利的大会 / Victory Meeting
版本：v1.0.0-test（2026-08-19 首发）
平台：macOS Apple Silicon（M 系列芯片）
下载：GitHub Release → VictoryMeeting-1.0.0-arm64.dmg

---

一、它做什么

会议开完，真正重要的东西却没有留下来——结论、任务、风险点、客户承诺，散落在群聊、录音、个人备忘里，没有形成可复用的资产。

胜利的大会要解决的就是这件事：把一次会议的录音，变成可操作的业务文件 + 可积累的知识资产。

一句话：「你的本地 AI 会议同事。」

---

二、核心流程

录音 / 上传音频
       ↓
   本地转写
       ↓
深度业务纪要（结论 · 重点 · 行动项 · 风险 · 下一步）
       ↓
   待办事项提取
       ↓
  候选知识发现
       ↓
   人工确认
       ↓
正式写入知识库

每一步都有明确的产出物，不是一张模糊的「AI 摘要」。候选知识必须经过人工确认才能进入知识库——这是防止噪音污染的关键设计。

---

三、功能清单

功能	说明
会议录音	支持直接录音，文件优先存本机
上传录音	上传已有音频，生成转写和纪要
本地转写	本地模型转写，不上传录音到云端
深度业务纪要	按场景整理：结论、重点、行动项、风险、下一步
待办事项	从会议承诺提取明确、可执行的任务
会议库	统一管理历史会议、转写、纪要、附件
知识库	沉淀客户/产品/渠道/项目/竞品/政策/方法论
人工确认	候选知识审核后才写入正式知识库
多格式导出	导出为常用文档格式，方便同步给团队
Obsidian 集成	可选：知识库与 Obsidian 联动

---

四、适合哪些场景

产品的定位非常具体——面向有大量客户、渠道、代理商沟通的业务团队，而不是泛用型会议工具：

连锁谈判：沉淀客户诉求、价格策略、资源承诺、风险点和下一步动作。

代理商沟通：记录区域问题、政策执行进度、反馈和需要总部支持的事项。

产品方案评审：整理产品定位、卖点、价格、渠道策略和上市节奏。

市场周会：复盘目标达成、费用使用、活动进展和下周计划。

项目复盘：保留关键问题、根因、有效动作和可复用方法论。

这些场景的共同特征：会议信息有商业价值，但当前的沉淀方式很差。大量知识存在个人脑子里、聊天记录里，没有系统化。

---

五、数据边界

• 录音、转写、纪要和知识库数据优先保存在本机
• API Key 不写入代码，不应公开分享
• 配置线上 AI（DeepSeek 或兼容接口）后，只在生成纪要或知识发现时调用——不持续上传数据
• 未经人工确认的候选知识不会进入正式知识库

---

六、安装

当前版本仅支持 macOS Apple Silicon（M 系列），Intel Mac、Windows、Linux 暂不支持。

1. 到 GitHub Release 下载 VictoryMeeting-1.0.0-arm64.dmg
2. 安装，首次打开按引导完成：
   · 检查本地转写环境
   · 配置 DeepSeek 或其他兼容 API
   · 做一次 20 秒测试录音
   · 按需配置 Obsidian 知识库

---

七、这个产品在解决什么

中国大量企业的「业务知识沉淀」问题是真实的：渠道经理换人、区域负责人离职、老客户关系断裂，重要的背景信息随着人走了。会议里谈的东西从来没有被系统化保留。

胜利的大会的切入点很准——不是做一个通用型会议录制工具，而是做一个业务场景下的知识沉淀系统，AI 做的是把录音里的业务信息提炼成结构化资产，人负责最后的确认和判断。

v1.0.0-test 是第一个公开测试版，仅 macOS Apple Silicon，功能还处于早期阶段。值得跟踪。

---

Mycelium Protocol — 追踪 AI 系统的底层演化

---

关于 Mycelium

菌丝协议。持续追踪 AI 工具、系统和实验的内容节点。

---

Victory Meeting (胜利的大会): Local AI Meeting Assistant — Recording → Business Notes → Knowledge Base, macOS First Release

by Mycelium Protocol

---

GitHub: zhangchunquan298-anhui/shengli-dahui
Name: 胜利的大会 / Victory Meeting
Version: v1.0.0-test (released 2026-08-19)
Platform: macOS Apple Silicon (M-series) only
Download: GitHub Release → VictoryMeeting-1.0.0-arm64.dmg

---

What It Does

After meetings end, the things that actually mattered — conclusions, task assignments, risk flags, customer commitments — scatter across chat threads, audio files, and personal notes. They don’t accumulate into reusable assets.

Victory Meeting addresses this by turning a meeting recording into actionable business documents + accumulated knowledge assets.

Tagline: “Your local AI meeting coworker.”

---

Core Workflow

Record / upload audio
         ↓
  Local transcription
         ↓
Deep business notes (conclusions · key points · actions · risks · next steps)
         ↓
   Action-item extraction
         ↓
  Candidate knowledge discovery
         ↓
    Human confirmation
         ↓
 Written into knowledge base

Each step produces a concrete output — not a generic “AI summary.” Candidate knowledge requires human review before entering the knowledge base — the core safeguard against noise pollution.

---

Features

Feature	Description
Recording	Record meetings directly; files stored locally
Audio upload	Upload existing audio for transcription and notes
Local transcription	Runs locally; recordings not sent to cloud
Business notes	Structured output: conclusions, key points, actions, risks, next steps
Action items	Clear and executable tasks extracted from commitments
Meeting library	Manage meeting history, transcripts, notes, files
Knowledge base	Customer / product / channel / project / competitor / methodology knowledge
Human confirmation	Candidate knowledge reviewed before saved
Export	Common document formats for team sharing
Obsidian integration	Optional: sync knowledge base with Obsidian

---

Target Scenarios

The product is specifically positioned for business teams with high volumes of client, channel, and distributor meetings — not generic meeting software:

Retail-chain negotiation: capture customer needs, pricing logic, resource commitments, risks, and follow-ups.

Distributor meetings: record regional issues, policy execution, progress updates, and support requests.

Product reviews: summarize positioning, selling points, pricing, channel strategy, and launch rhythm.

Market weekly meetings: review targets, spending, activities, and next-week plans.

Project retrospectives: retain issues, root causes, effective actions, and reusable methods.

Common thread across these scenarios: meeting content has business value, but current retention methods are poor. Most knowledge lives in individual memories or chat logs with no systematic structure.

---

Data Boundary

• Recordings, transcripts, notes, and knowledge data stored locally first
• API keys not hard-coded; should not be shared publicly
• AI API (DeepSeek or compatible) called only for note generation and knowledge discovery — not for continuous data upload
• Candidate knowledge requires human confirmation before entering the formal knowledge base

---

Install

Currently macOS Apple Silicon (M-series) only. Intel Mac, Windows, Linux not yet supported.

1. Download VictoryMeeting-1.0.0-arm64.dmg from GitHub Releases
2. Install and complete the first-run guide:
   · Check local transcription environment
   · Configure DeepSeek or compatible AI API
   · Do a 20-second test recording
   · Optional: set up Obsidian knowledge base

---

What Problem This Is Really Solving

In a large number of Chinese enterprises, “business knowledge retention” is a genuine problem: when a channel manager or regional lead leaves, critical relationship context goes with them. What was agreed in meetings has never been systematically preserved.

Victory Meeting’s positioning is specific — not a general-purpose meeting recorder, but a knowledge accumulation system for business contexts, where AI handles the extraction of structured assets from recordings and humans handle the final review.

v1.0.0-test is the first public release, macOS Apple Silicon only, early stage. Worth tracking.

---

Mycelium Protocol — tracking the deep evolution of AI systems

© 2026 Mycelium Protocol. All rights reserved.